DISCUSSION PAPER SERIES

IZA DP No. 12099

Agata MaidaAndrea Weber

Female Leadership and Gender Gap within Firms: Evidence from an Italian Board Reform

JANUARY 2019

Any opinions expressed in this paper are those of the author(s) and not those of IZA. Research published in this series may include views on policy, but IZA takes no institutional policy positions. The IZA research network is committed to the IZA Guiding Principles of Research Integrity.The IZA Institute of Labor Economics is an independent economic research institute that conducts research in labor economics and offers evidence-based policy advice on labor market issues. Supported by the Deutsche Post Foundation, IZA runs the world’s largest network of economists, whose research aims to provide answers to the global labor market challenges of our time. Our key objective is to build bridges between academic research, policymakers and society.IZA Discussion Papers often represent preliminary work and are circulated to encourage discussion. Citation of such a paper should account for its provisional character. A revised version may be available directly from the author.

Schaumburg-Lippe-Straße 5–953113 Bonn, Germany

Phone: +49-228-3894-0Email: publications@iza.org www.iza.org

IZA – Institute of Labor Economics

DISCUSSION PAPER SERIES

IZA DP No. 12099

Female Leadership and Gender Gap within Firms: Evidence from an Italian Board Reform

JANUARY 2019

Agata MaidaUniversità di Milano and LABORatorioRevelli, Collegio Carlo Alberto

Andrea WeberCEU, WU Vienna, IZA and CEPR

ABSTRACT

IZA DP No. 12099 JANUARY 2019

Female Leadership and Gender Gap within Firms: Evidence from an Italian Board Reform*

Over the last decade, several countries have followed the Norwegian example and

introduced laws mandating gender quota for corporate board membership. The main

aim of these laws is breaking the “glass ceiling” which prevents women from advancing

into top corporate positions. In this paper, we evaluate the Italian law of 2011, which

installed a step-wise increase in gender quota that remain effective for three consecutive

board renewals of listed limited liability firms. We link firm-level information on board

membership and board election dates with detailed employment and earnings records

from the Social Security registers. Exploiting the staggered introduction of the gender quota

regulation and variation in board renewals across firms, we evaluate the effect of the board

gender composition on measures of gender diversity in top positions over a period of 4

years. While the reform substantially raised the female membership on corporate boards,

we find no evidence of spillover effects on the representation of women in top executive

or top earnings positions. Our results confirm the findings by Bertrand et al. (2019) who

study the introduction of a gender quota for board members in Norway. Given that Italy

is a much less egalitarian society than Norway, with a larger scope of establishing gender

equality, our results confirm that board quota policies alone are ineffective in raising female

representation in top corporate positions, at least in the short run.

JEL Classification: J24, J7, J78

Keywords: gender quota, corporate board reform, glass ceiling, female employment

Corresponding author:Andrea WeberCentral European UniversityDepartment of EconomicsNádor utca 91051 BudapestHungary

E-mail: WeberA@ceu.edu

* We benefitted from discussions with Tito Boeri, David Card, Pietro Garibaldi, Astrid Kunze and Josef Zweimüller.

We thank seminar and conference participants at Italian Social Security Institute (INPS), Research Institute for the

Evaluation of Public Policies (IRVAPP, Trento) Department of Econometrics, Statistics and Applied Economy of

University of Barcelona, ESPE-Antwerpen, EEA-Cologne, EALE-Lyon, Brucchi Luchino workshop –Rome. For providing

invaluable support with the data access, we thank the research division of the Social Security Institute of Italy (INPS),

especially Massimo Antichi, Mariella Cozzolino, Edoardo Di Porto, and Paolo Naticchioni. We gratefully acknowledge

funding and data ac-cess through the VisitINPS Scholars A programme. The usual disclaimers apply.

2

Introduction

The low representation of women in top earnings positions has been identified as a major obstacle for closing the

remaining gender pay gaps and achieving full gender equality (Fortin et al., 2017). Even though women make up

almost half of the labor force they are becoming increasingly scarce the higher one moves in the earnings

distribution and they are also severely under-represented in leadership positions. Mandated gender quota have

been suggested as a policy measure to promote female career progression towards the top. In 2003 Norway was

the first country to pass a law requiring a minimum of representation of 40% for each gender on the board of

directors of publicly limited companies. Since then, Austria, Belgium, Denmark, France, Ireland, Iceland, Italy,

Germany, the Netherlands and Spain have followed with similar regulations. In the Europe 2020 Strategy the

European Commission proposes a law that requires a minimum of 40% female board members in listed

companies across the European Union.

Gender quota laws are generally highly successful in raising the number of female members on corporate boards.

By 2016 female board membership in large listed companies in the EU has increased to 23% on average, up

from only 12% in 2010. The biggest changes have occurred in countries that have introduced a female quota law

in recent years (European Union, 2016). However, it is much less clear whether gender quota laws also have an

impact on women outside corporate boards. Matsa and Miller (2011) show that US companies with higher shares

of female board members are more likely to hire female top executives. But it is not entirely clear whether this is

a causal link or whether unobserved factors are driving both outcomes. In their seminal study, Bertrand et al.

(2019) exploit quasi-experimental variation from the Norwegian law and find no evidence that the quota

regulation benefited other women employed in companies subject to the quota, neither do they find an impact on

highly qualified women.

In this paper, we evaluate short run effects of the Italian law of 2011 that introduced a gradual and temporary

gender quota for boards of directors and auditors of companies listed at the Italian Stock Exchange. In contrast to

Norway, which is considered to be one of the most gender equal countries in the world, Italy is characterized by

a highly conservative gender culture. In the Global Gender Gap Report 20171 that benchmarks 144 countries,

Norway ranks second while Italy is far behind on rank 82. It is thus interesting to see whether gender quota law

is potentially more powerful in a conservative, Southern European culture.

As our data cover at most four post-reform years, we expect that the most immediate impact of the change in the

board composition manifests itself in decisions concerning the gender of employees in top management position,

before it starts trickling down to lower levels in the hierarchy. Therefore, our main focus is on gender diversity

in top executive and high earnings positions in companies that are subject to the law. Our empirical analysis is

based on detailed data that match company-level information on board composition and board elections to

1 https://www.weforum.org/reports/the-global-gender-gap-report-2017

3

administrative social security record records which allow us to compute annual information on gender workforce

composition and top executive positions for the universe of Italian private sector firms over the period 2008 –

2016.

To identify the effects of the mandated change in board composition on gender disparities within companies, we

pursue three different strategies. First, we exploit variation in the timing of board elections at which gender

quota become binding in listed firms. Second, we compare post-election outcomes in listed firms with a matched

comparison group of limited non-listed firms that are not subject to the law applying a difference-in-difference

design. Third, we consider heterogeneity in the incentive for board adjustments at the company level by

exploiting variation in the share of female board members in the pre-reform period.

The annual reports of the regulatory board of the Italian stock exchange which supervises compliance with the

law (CONSOB 2016, 2017) as well as a recent paper by Ferrari et al. (2018) document large and significant

increases in the share of female board members in listed companies. In addition, we observe that some

companies even acted in anticipation of the law and increased gender diversity on their boards in 2012 board

elections before the law was binding. We also show, that in the pre-reform years the share of female board

members was strongly correlated with the representation of women in the workforce and in highly qualified

positions which is in line with the findings by Matsa and Miller (2011).

Our estimation results point towards the following main findings. First, our results provide no evidence that the

female board quota mandated by the Italian law led to an increase in female representation at the top executive

level or among top earners, at least not in the short run. Second, there is some indication that listed companies

promoted one of their female managers as a CEO. But these promotions did not increase the representation of

women among top earners in the company. Heterogeneity analysis by the pre-reform share of female board

members reveals that new appointments of female CEOs occurred only in companies that already fulfilled the

female quota in 2012. Third, when we extend the focus of our analysis to the wider group of all female

employees at listed companies or the implementation of family friendly policies, there is again no evidence of

changes in the overall gender workforce composition due to the reform.

Our evidence for Italy thus fully confirms the findings for Norway by Bertrand et al. (2019). Even though both

countries differ strongly in terms of gender attitudes and female labor market outcomes, the strict enforcement of

gender ratios on boards of directors are equally ineffective in improving female career progression towards the

top.

The rest of the paper is organized as follows. Section 1 presents a review of the recent literature and highlights

the contribution of our paper. Section 2 explains the Italian institutional background and the Golfo-Mosca law.

Section 3 describes data and presents descriptive evidence. Section 4 discusses the empirical methodology and

4

provides justification for our identification assumptions. Section 5 presents estimation results and Section 6

concludes.

1. Literature Review

In most countries, the introduction of gender board quota policies aroused skepticism and was accompanied by

vivid debates about the potential effects on firm performance and the economy as a whole. Proponents

emphasized the need to increase gender diversity in top positions and the importance of providing female role

models. In addition, experience with female board members should correct negative perceptions about women in

leadership positions and reduce statistical discrimination. Opponents were worried that the legal restriction of the

company’s optimal choices of board members and an insufficient supply of qualified female candidates for

board positions might harm business outcomes. While the different positions were widely discussed the

empirical evidence is surprisingly inconclusive.

The corporate governance literature provides credible evidence that managers’ personal traits influence corporate

decisions (Malmendier et al. 2011, Bertrand and Schoar (2003)). However, it is less clear whether gender

diversity among executives matters directly. At the end of the fierce selection process women reaching top

executive positions have vastly different characteristics and preferences than women in the average population

and are more similar to men in the same positions (Adams and Funk, 2012). Furthermore, it is unclear whether

females in higher ranks of the hierarchy actively promote other women’s careers (Staines, Tavris, & Jayarante,

1974). Kunze and Miller (2017) find evidence for positive spillovers of females across ranks in large Norwegian

companies. But Bagues et al. (2017) find evidence for the opposite behavior. They document that female

evaluators are less favorable toward female candidates than toward males, such that additional female evaluators

effectively worsen the chances of female candidates.

Multiple empirical studies document a positive relationship between female leadership and various female labor

market outcomes at the firm level, among them female employment, gender wage gaps, retention rates after

economic shocks, or the use of flexible employment contracts such as part-time work (Flabbi et al., 2018; Tate

and Yang, 2015; Cardoso and Winter-Ebmer, 2010; Gagliarducci and Paserman, 2015; Matsa & Miller, 2014;

Lucifora and Vigani, 2016; Devicienti et al., 2018). These results are either based on cross-sectional evidence or

on models with time invariant firm fixed effects that do not necessarily have a straight forward causal

interpretation.

Another strand of the literature focuses on the impacts of gender diversity on corporate boards on firm outcomes.

In a sample of US firms Adams and Ferreira (2009) find that female directors behave differently than males,

they have higher attendance at board meetings and gender-diverse boards spend more effort on monitoring.

5

However, the relationship between board diversity and firm performance is ambiguous and depends on the

strength of governance.

Smith (2018) and Ferreira (2015) provide comprehensive surveys of the literature studying the relationship

between board diversity and firm performance. As Norway was the first country to enforce a mandatory gender

quota policy a decade ago, the majority of studies focus on this reform. But the evidence remains surprisingly

inconclusive. Early studies by Ahern and Dittmar (2012) and Matsa and Miller (2013) find rather negative short-

term effects on firm value and Tobin’s Q, labor costs and short run profits, while more recent studies draw a

more nuanced picture. Evidence from other countries shows a similarly vague picture. Comi et al. (2016)

examine the effects of mandated gender quotas on firm performance in four EU countries: Belgium, France, Italy

and Spain and report mixed results. Flabbi et al. (2017) document that companies with female board members in

Latin America and the Caribbean have a higher probability to appoint female executives. Ferrari et al. (2018)

study the short run effects of the Italian reform and find positive responses to board renewals in stock market

values. Overall, it seems that the success of mandated board quotas depends on the institutional and legal

framework in the country as well as the governance structure within companies, which poses serious challenges

for the design of efficient policies.

While the question how board quota mandates affect firm performance has received a lot of attention in

empirical research, the effects on gender discrimination and female labor market outcomes have not been studied

as extensively (see Matsa and Miller, 2011 for the US; and Bertrand et al., 2019 for Norway). Our paper

contributes evidence from Italy to the question whether smashing the glass ceiling at the board level can reduce

gender discrimination at lower levels in the company.

2. The Italian Gender Quota Law and Board Composition

Following the example of other countries, Italy introduced a gender quota for the members of boards of directors

in companies listed with the Italian stock exchange. The so-called Golfo-Mosca2 law was passed in parliament in

June 2011 and went into effect on August 12, 2012. But unlike in other countries, the Italian law imposes a time

limited measure that is gradually introduced. It aims at supporting a cultural renovation and removal of barriers

that limit the access of women to leading positions.

In Italy, each listed firm appoints the board of directors via elections that are typically held every three years in

the period from March to June. The Golfo-Mosca law requests a minimum of 1/5 of board seats for each gender

with the first board appointment following August 2012, and a minimum of 1/3 starting with the second

2 The law 120/2011 is named after Golfo and Mosca, who were the two first members of parliament, who signed it.

6

appointment.3 The quota regulation mandated by the law expires with the third term of board appointments. In

contrast to the widely studied Norwegian law, which was passed in 2003 but not strictly enforced before 2008,

the timing in the Italian case is sharper, which is an advantage for identification of the reform effects.

The regulatory board of the Italian stock exchange, CONSOB, is in charge of monitoring and supervising

compliance with the law.4 In case of noncompliance, CONSOB first issues warnings with 60 day deadlines. If

the initial deadlines expire, administrative penalties between €10.000 and €100.000 can be administered and the

company is given another three months to comply with the regulation. Finally, if non-compliance persists, the

elected board will lose legitimacy.

In the annual report, CONSOB documents female representation on corporate boards of Italian listed companies

(CONSOB 2016, 2017). Figure 1, based on CONSOB data shows the development of the share of females

among board members in Italian listed firms over time. In 2008 the average company started out with a very low

female board share of 5.9 percent which was rising slowly over the subsequent years. But from 2012 onwards,

we see a strong increase in the female shares to levels which are in line with the introduction of the quota

regulation that applied to additional cohorts of companies holding board elections over the years. The gender

quota of 20 percent stipulated in the first round of board appointments was already surpassed in 2014 and in June

2017 the overall share of female board members was about one third, which implies that some boards appointed

more female members than required by the law. Table 1 further confirms the high compliance with the law;

while in 2008 only 43% of companies had at least one female board member, male-only boards were virtually

eliminated by 2015 with the first round of board appointments (see column (4)). The average board of directors

has 10 members and the reform increased the number of positions held by females from 170 in 2008 to 758 in

2017. CONSOB further reports that the Golfo-Mosca law increased board diversity along other dimensions.

With the appointments of more female directors, the average age of board members has declined, board

members have on average higher educational levels, and they include more non-Italians and fewer family

members.

It is not uncommon for directors to hold seats on multiple boards, which might lead to concerns that a select

group of few women hold all the female board seats. According to CONSOB (2017) the average number of seats

held by a female director is slightly higher than that of a male director. Between 2013 and 2017 number of

female board seats has increased by 80% and the number of individual women holding board seats has increased

by 51%, which implies that more women are taking a second seat. Conversely male board seats have declined by

23% but the number of individual men who are on boards declined only by 20% as some of them lost a second

3 The decimals arising from application of one-fifth and one-third are rounded up the nearest integer. 4 A company must communicate their board composition within 15 days of renewal and/or substitution of members. The administrative, management and supervisory bodies must communicate possible non-compliance with gender quota.

7

seat.5 Overall, the changes in board composition along with the swift compliance with the gender quota suggest

that the male dominance on boards of directors prior to the law was not driven by a shortage in the supply of

women qualified for director positions (Ferrari et. al., 2018).

3. Data and Descriptive Analysis

We construct our dataset by linking information about boards of directors and board renewal dates of companies

listed with the Italian stock exchange with longitudinal administrative firm-worker records provided by Italian

Social Security Institute (INPS).6

We start with the names of 241 companies listed with the Italian stock exchange in 2013, who are subject to the

Golfo-Mosca law according to the CONSOB website. From this website we also downloaded the names of each

company’s board members. We identify the gender of board members from their first names.7 In addition, we

manually collected the election date of the board of directors from each company’s website. We further restrict

the sample to companies listed in the stock market 2013, who entered in the listed marked before the reform and

remained listed until December 2016. This sample of firms continuously listed between 2012 and 2016 includes

200 firms. One concern is that with the introduction of the Golfo-Mosca law, some companies might delist from

the stock exchange in order to avoid the legal requirement. But according to historical data directly available at

the official website of the Borsa Italiana (Italian Stock Market), Italy didn’t register a peak in the rate of

delistings after the announcement of the Golfo Mosca law8 (Borsa Italiana, 2018).

In the next step, company names are merged to the administrative records from INPS archives, which cover the

entire population of dependent workers’ job spells in the private non- agricultural sector of the Italian economy

and include detailed information on about 18 million workers and 1.5 million firms per year. The match rate of

this merge is 99.6%. From the INPS data, we construct yearly variables with firm-level information on

workforce characteristics and executives from 2008 to 2016. In our empirical analysis, we focus on a balanced

panel of companies who are continuously observed in INPS (i.e. employing a positive number of workers each

year) over the full period 2008-2016. This sample includes 188 listed companies.

From the information on the universe of workers employed in each company, we derive annual workforce

characteristics capturing the distribution of workers by gender and occupation and the within firm wage

distribution.9 Furthermore, by following the inflow and outflow of workers into a company over the year, we

5 For female interlocking on corporate boards of Italian listed companies see CONSOB (2017), Table 2.21. 6 Data access is available through VisitINPS program https://www.inps.it/nuovoportaleinps/default.aspx?itemDir=47212 7In order to identify the few ambiguous cases, we checked on line (i.e. website like linkedin) 8 https://www.borsaitaliana.it/borsaitaliana/statistiche/mediaitaliano/statistiche/mercatoprimario/2018/revoche.en_pdf.htm 9 For the construction of firm-level variables, we exclude workers younger than 15 and older than 64 and workers who worked less than eight weeks in a given year.

8

compute the number of hires and layoffs by gender and year. In addition, the INPS data records managers as a

separate occupational category, which we exploit to construct information on executives. We use the following

variables capturing gender diversity in leadership positions: an indicator variable equal to one if one of the

managers is female, the share of workers who are female managers, and an indicator equal to one if the CEO is

female. In the definition of the CEO we follow Flabbi et al. (2018) who identify the CEO as the manager with

the highest wage in a given firm-year. We also consider the representation of females in the upper part of the

firm-specific wage distribution and compute the share of female workers who are have wage earnings in the top

quartile and the top decile of the distribution, respectively.

Since the law affects only companies which are listed with the stock exchange, we can construct a comparison

group of no-listed firms. As INPS data provide information about the legal form of the companies, we select the

comparison from the sample of limited companies who employ at least one manager and have continuous INPS

records from 2008 to 2016.10 Descriptive statistics for the samples of listed and limited companies in 2012 are

shown in Table 2 columns (1) and (2). The two groups of companies differ markedly in terms of the firm size

distribution and by the occupational distribution. Listed firms having a stronger focus on more highly qualified

white collar workers and managers than the average limited company. Accordingly, average wage levels are also

significantly higher among listed firms. There is also a gap in female employment between listed firms, who

employ 40% female workers on average and limited firms with an average female workforce of 32%. Female

representation in executive and high earnings positions is low in all firms. But listed firms are significantly more

likely to employ at least one female manager.

To reduce differences in observed pre-reform characteristic between the group of listed firms and the

comparison group we apply a matching procedure. In particular, we estimate a propensity score for the

probability that the firm is subject to the law based on observed workforce characteristics in 2008 – 2012 along

with a detailed set of regional (42 provinces) and industry (46 industry codes at two digit level according to

ATECO07) indicators. Figure A1 in the Online Appendix shows the distribution of predicted propensity scores

for the samples of listed companies and limited non-listed companies in Table 2 columns (1) and (2). We select

the matched comparison group of non-listed companies via nearest neighbor matching using a Mahalanobis

matching metric based on several key variables.11 In order to improve the quality of the match, we drop

observations from either group of companies if their number of employees is in the top or the bottom 1

percentiles of the annual distribution in any year from 2008 to 2012. This reduces the number of companies in

the balanced panel to 167 listed companies and 149 matched control companies. The t-tests for the mean

difference between listed and matched control firms of almost all variables included in the matching procedure

are not statistically different from zero (see Table A1 in the Online Appendix).

10 Over the period of observation there are 23,747 limited firms in the data and 7770 limited firms with at least one manager. 11 The list of variables included in the matching metric and in the propensity score estimation can be found in the Appendix.

9

Table 2, columns (3) and (4), reports summary statistics of key variables in the pre-reform year 2012 for the

matched samples which confirm that the main differences between listed and matched non-listed firms have

disappeared. The average firm employs about 630 workers of which about 40% are female. The average gross

earnings of full-time workers are 1000 Euro per month and about 7% of workers are employed part-time, most

of them are female. About 20% of the workforce are blue-collar workers and 70% are white collar. About 7% of

workers in listed firms are managers. About one quarter of employees in the pool of managers are female. While

about 70% of the listed and matched limited companies have at least one female manager, only 5% have a

female CEO.12 Women are also strongly underrepresented in the top of the firm-specific wage distribution. About

20% of workers in the top quartile of the distribution are female and only ten in 100 workers with earnings in the

top decile of the firm-specific wage distribution is female.13 These statistics imply that in 2012 Italian listed

companies were far away from gender equality in top executive positions and in top earnings positions.

The panels in Figure 2 compare the evolution of our gender diversity measures in the matched sample of listed

and non-listed firms before and after the introduction of the Golfo-Mosca law in 2012. As shown in the graphs,

female representation in leadership positions mostly follows rising trends over time in both groups of companies,

but it remains very low over the whole period. The share of firms with at least one female manager in Panel A

has risen by about 5 percentage points from 70% in 2008 to 75% in 2016 among listed firms and from 57% to

62% in limited non-listed firms. The percentage of listed firms with a female CEO, Panel C, increased from

about 6% in 2008 to 8% in 2016, while there was no corresponding increase in the share of firms with a female

CEO among non-listed firms. We also see that female representation has increased in the top part of the wage

distribution, Panels D and E. But as it started from very low levels, the increase over time did not achieve much

in terms of gender equality. Perhaps most importantly, the figures do not show major changes in the gaps

between the listed and non-listed firms from 2012 onward, which would be indicative of the effect of the board

gender quota on diversity outcomes.

To confirm the visual impression from the figures, we estimate for each outcome measure 𝑌"#a linear regression

model with a full set of firm dummies 𝛼", year dummies 𝛾# , and a set of year dummies interacted with an

indicator variable 𝐿" equal to one for listed firms as denoted in the following equation

𝑌"# = ( 𝛿*𝐿"Ι(𝑡 =.

*/0.,*23

2012 + 𝑘) + 𝛾#+𝛼" + 𝜖"#(1)

The model parameters δ= measure the difference between listed and non-listed firms over time. We report

parameter estimates in Online Appendix Table A2 and plot δ>= Y@A3BA⁄ , the coefficient estimate relative to the 12 Flabbi et al. (2018) study a sample of Italian manufacturing companies in the 1990’s and report lower shares of female workers (26%), female executives (3%) and female CEO’s (2%). In their sample only 20% of firms have at least one female executive in 1997. 13 Table 2 reports the mean of the share of workers who are female and have earnings in the top quartile or decile. To get the share of female workers in the top quartile or decile we multiply the means by 4 and 10, respectively.

10

mean value of Y among listed firms in 2012 for selected variables in Figure 3. The results confirm that

differences between listed firms and matched control firms are small and statistically insignificant prior to 2012

which implies that our matched control sample is a valid comparison group for the listed firms in the pre-reform

period. After 2012, we do not see an increase in the outcome gap between listed and non-listed firms which

could be attributed to the rising share of female board members in Figure 1. For most outcomes the change

between 2012 and 2016 is less than 10% and none of them is statistically significant, except the share of part-

time workers in 2016. In Figure 3, panel B we see that changes in the share of firms with a female CEO are

particularly noisy. This is due to the low probability of observing a female CEO and the relatively small sample

size. In 2012 only 8 listed firms have a female CEO. This implies that if one additional firm hires a female CEO

in the next year this already corresponds to a change by 12%. In the next section, we exploit the exact timing of

the quota regulation at the firm level to estimate the magnitude of the reform effects.

4. Empirical Strategy

The Golfo-Mosca law links the mandated gender quota to board elections which are typically held every three

years. Election years vary across companies and they are arguably exogenous with respect to the introduction of

the reform. Thus, the law creates variation in the timing when the quota rule becomes binding at the firm level.

In 2016, the last year of our data, all firms should have completed the first round of board appointments at which

the mandated minimum share of each gender of 20 percent on the board of directors applies.14 Therefore our

estimates can be interpreted as reflecting the short-run effects of the first step of the Golfo-Mosca reform.

Our first empirical strategy combines three different types of variation for identification of the effect of female

board representation on gender diversity within the company. First, we exploit the variation in the timing of the

board election among the set of listed firms. Second, we include the matched comparison companies who are

unaffected by the reform and perform a difference-in-difference analysis. Our third strategy additionally

considers heterogeneity in the incentive for board adjustments generated by the reform at the company level by

exploiting variation in the share of female board members in the pre-reform period.

Table 3 shows descriptive statistics by the three different cohorts of companies with the first board renewal in

2013, 2014, and 2015, respectively. Roughly on third of our sample of matched listed companies have scheduled

board elections in each of the three years. Companies with elections in 2014 tend to be slightly larger, employ

more blue-collar workers and less females and are also less likely to have a female manager than those in the

other two board election cohorts. But due to the small sample sizes the differences are mostly insignificant,

which confirms our assumption that the board election years are as good as randomly assigned. However, if we

14 Only the group of firms who had their first board elections after the implementation of the law in 2013 have reached the second step of the gender quota by 2016 and are required to reach a female share of one third of their board members.

11

look at the shares of female directors, it becomes apparent that anticipation of the Golfo-Mosca law already

influenced board elections before it became binding. The share of female board members who are in office in

2012 increases over election cohorts. Companies with the most recently elected boards, i.e. those in election

cohort 2015, already have an average share of 13.6% of female directors in 2012, while companies with

scheduled renewals in 2013 on average only have 8.6% female directors in 2012. 15 In the pre-reform year less

than a fifth of companies in the 2013 and 2014 election cohorts fulfill the 20% female board quota, but about one

third of the companies in the 2015 election cohort do. These companies will be defined as “low incentive” in the

next section. The anticipation pattern also persists after the first board appointment following the reform. While

firms in the 2013 election cohort employ 25% female directors after the first board renewal, firms in the 2015

cohort employ 29% and are thus closer to the final quota threshold than to the threshold set by first reform step.

Our first identification strategy exploiting variation in the timing of board elections estimates of the following

model for the set of listed firms:

𝑌"# = 𝛽B𝐷"# + 𝛽A𝑋"# + 𝛾#+𝛼" + 𝜖"#(2)

where Yit is a variable representing one of our gender diversity measures for firm i in year t. Dit is a dummy

variable that is equal to zero in the years prior to firm i’s first board appointment and 1 in the appointment year

and in all subsequent years. β1 is the main parameter of interest capturing the average change of the outcomes

associated with the reform. 𝛼" is a firm fixed effect that captures any time invariant unobserved firm

characteristics and 𝛾# represent a set of year dummies. The vector Xit represents controls for firm size and firm

size squared.

The identification strategy in equation (2) compares outcomes in firms who appoint a new board subject to the

gender quota early with firms who are scheduled to appoint their board later. Because boards are reappointed

every three years, the time window between renewals in the first and the third cohort of firms is only 2 years. For

example, we compare outcomes in year 2013 between firms that renew their boards in 2013 and firms that renew

in 2014 and we compare outcomes in 2013 and 2014 between firms that renew their boards in 2013 and firms

that renew in 2015. So the estimates of equation (2) only give us the immediate or very short run effects of the

reform on firm outcomes.

To get at the longer-run effects, we estimate a model for listed and limited non-listed firms in the matched

comparison group:

𝑌"# = 𝛽B𝐿" ∗ 𝐷"# + 𝛽A𝑋"# + 𝛾#+𝛼" + 𝜖"# (3)

15 Companies that held board elections in January to May 2012 are included in the sample. But we removed three firms that renewed their board the in the last months of 2012. Including these firms in the sample does not change the summary statistics of the final sample nor the econometric results.

12

In addition to the comparison within listed firms, this specification adds a difference-in-difference design across

both groups of companies. If a change in the gender composition of the board of directors affects firm level

outcomes with some delay, we should see larger estimates in in the model with the full set of companies than in

the model only including listed firms.

The estimation sample consists of a balanced panel of firms observed in each year from 2008 to 2016.

Observations are weighted by firm size to make the analysis representative of the economy wide impact. To

control for potential autocorrelation in the error term, standard errors are clustered at the firm level.

We have discussed that the average share of women on boards of directors of listed firms in Italy was extremely

low before the introduction of the Golfo-Mosca law; in 2011 only 7.4% of all board members were female.

Nevertheless, there was heterogeneity in board gender diversity across firms with some firms being close to the

threshold already before the law was passed and others that faced a wide margin of adjustment, for example if it

had appointed male-only boards before 2012. With the implementation of the Golfo-Mosca law, companies thus

faced different incentives to adjust their board gender composition. Our third identification strategy exploits this

heterogeneity by estimating separate reform effects for different groups of companies.16 In particular, we

consider three groups of companies depending on their board gender composition in 2012.17 The first group of

firms facing a “high incentive” to change the board composition had less than 10% female board members in

2012. This group includes 47% of the listed companies in our matched sample. The second group facing a

“medium incentive” for adjustment had more than 10% but less than 20% of female board members in 2012 and

it consists of 31% of listed companies. The third, “low incentive”, group consists of firms who already fulfilled

the first step of the quota regulation and have 20% or more female board members in 2012. This group includes

22% of our sample.18

Table 4 shows descriptive statistics of firm characteristics in year 2012 for each group of companies. The table

shows that gender diversity at the firm level is correlated with diversity on the board of directors. Low incentive

firms that already fulfill the gender quota in 2012 have on average higher shares of female employees and also

higher shares of women at the top of the earnings distribution. In addition, those firms have higher shares of

female managers. No firm in the high incentive group has a female CEOs. This is not particularly surprising,

because the CEO is typically a member of the board of directors. In addition, as remarked above, gender

diversity at the board in 2012 is also correlated with the election cohort due to anticipation of the female quota

policy.

16 This identification strategy resembles Bertrand et al. (2019) who instrument the contemporaneous pre-reform female board share with the pre-reform female board share interacted with year dummies. 17 Currently we do not have data on the board composition in 2011, the year before the Golfo-Mosca law was passed. 2011 would be a better baseline year heterogeneity analysis, because there was less anticipation of the reform. 18 About half of the companies in the low incentive group are also in the 2015 board election cohort. This means that they reappointed their boards in 2012 and these elections were probably influenced by anticipation of the Golfo-Mosca law.

13

To test for differences in the reform effects across groups, we estimate the empirical models in equations (2) and

(3) interacting dummy variables for group membership with the reform indicator variable. In addition, we run a

separate matching procedure for each group of companies to select non-listed companies which are most similar

to group members in terms of observed characteristics.

4.1 Identification Assumptions and Validity Checks

Identification of the comparison of outcomes in listed and matched comparison companies relies on the

assumption that in the absence of the law mandating a change in the female share of board members, all outcome

variables in listed firms subject to the quota regulation would evolve in the same way as in the comparison firms.

While we cannot fully test the common trend assumption, we can check whether trends in outcomes in listed and

comparison firms developed in a similar way in the period before the law was introduced.

Our first validity check aims at testing for common pre-trends among listed firms who renew their boards early,

and those who renew their boards for the first time in later years. In particular, we perform a set of placebo to

tests, where we substitute the variable Dit in equation (2) with dummy variables, which turns equal to one in the

first and second year before the actual board election dates, respectively. The model specification including only

listed firms, in equation (2), effectively compares outcomes in the one or two year time windows between board

elections for different cohorts of listed firms. Thus bringing the placebo renewal year forward by one year should

reduce the estimated coefficient, because we only compare outcomes in the year of board renewal. Bringing the

placebo renewal date forward by two years only compares outcomes before any of the firms have changed their

boards and should thus result in estimated coefficients of zero. Online Appendix Tables A3 and A3_bis show

estimation results of this validity check. Indeed we find zero coefficients throughout for the placebo test which

shifts the board appointment date by two years.

The second validity check tests for differences in the pre-trends among listed companies and the matched sample

of limited non-listed companies. As explained above, our matching strategy takes annual values of all outcome

variables from 2008 to 2011 into account when estimating the propensity score. This should guarantee that both

groups of firms are fairly close in the pre-reform period, which is also confirmed by tests on the propensity score

shown in Online Appendix Table A1. In addition, we estimate the model in equation (1) which includes

coefficients for differences in outcomes at the yearly level. Table A2 in Online Appendix shows that none of the

coefficients is statistically significant in any of the pre-reform years.

These checks make us confident that the common trend assumption is valid in our application. There are further

concerns about identification, which we cannot test directly, however. First, firms with later scheduled dates of

board elections might respond to the Golfo-Mosca law in anticipation and already implement changes in gender

diversity policies before their boards are elected. If this is the case our first estimation strategy should produce

downward biased results. Second, companies might compete for highly qualified female workers who can be

14

promoted to top positions. If a listed firms have a higher incentive to hire these qualified women than non-listed

firms, we would expect an upward bias in our results.

A further identification assumption is that companies start responding to the Golfo-Mosca law in the year of the

first reappointment of their boards after 2012. We have seen in Figure 1 that the share of female board members

starts rising already in 2012 and in Table 3 we have shown that some companies who reappoint their boards in

2012 are influenced by anticipation of the reform. While the model assumes that companies with board elections

in 2015 only respond to the reform in 2015, anticipation might drive some of them already to respond in 2012. If

this is the case, our estimates might be downward biased. We therefore re-estimate our main specifications for

the sample that excludes companies with board renewals in 2015. This robustness check does not change the

main findings.19

5. Estimation Results

A central role of a company’s board of directors is to appoint and review company executives. If female

directors are more supportive of female executives, we might expect that gender diversity at the board has a

direct impact on the gender composition of the company’s top management. We therefore test if the Golfo-

Mosca law which mandated more diversity at the board level had an impact on diversity among company

executives. Estimation results for different measures of executive level gender diversity are shown in Table 5.

Panel A reports results for the model specification in equation (2), which focuses on listed firms only and

exploits variation in the timing of the gender quota mandate, and Panel B reports results for the full comparison

of listed and matched non-listed companies in equation (3).20 We have multiplied the coefficient estimates by 100

and can thus interpret the parameters as percentages.

We find that the share of companies that employ at least one female manager does not change due to the reform,

see column (1). Point estimates are very small and statistically insignificant. Also the share of female managers

in the overall workforce does not increase significantly as shown in column (2). The point estimate indicates at

most a positive increase of 0.04 percentage points (Panel A) from a mean of 1.6% in 2012. Estimates for whether

the firm has a female CEO in column (3) show a positive increase by one percentage point in Panel A. Given our

sample size of 167 listed firms this would imply that two additional firms appointed a female CEO at or

immediately after the first post-reform board election. Panel B shows that in the longer run comparison with the

matched control firms this effect becomes larger and statistically significant. Given that we see no change in the

overall share of female managers, this finding suggests that a more gender diverse boards are more likely to

19 Estimation results for the reduced sample are available on request. 20 Estimation results for specifications without covariates differ very little from the estimates in Table 5 and they are shown in Online Appendix Table A4.

15

promote one of the female managers to become the CEO. According to the estimate in Panel B, the fraction of

firms who have a female CEO increases by 3 percentage points, which means it almost doubles relative to the

base of 4.8% in 2012. Columns (4) and (5) in Table 5 show the impact on female representation in the upper part

of the firm specific earnings distribution. Our results point to a very small and positive increase in the share of

females in the top earnings quartile, which is significant only in the short run specification but not stable if we

consider the matched comparison group. There is no change at all in the share of females in the top earnings

decile. Thus even if firms become more likely to promote one of the company’s female managers as a CEO, this

is not accompanied by a substantial change in top female earnings. This result is at odds with Flabbi et al. (2018)

who report that female CEO’s have a positive impact on female wages in the top quartile of the female

distribution in large Italian manufacturing firms.

In Table 6, we investigate whether the reform led to changes in the overall gender composition of the workforce.

In particular, we study the share of females employed in the company and the female share in the inflow and

outflow of workers in columns (1) – (3). We also investigate whether firms with more gender diverse boards

invest more in family friendly workplace conditions proxied by the workforce shares of part-time workers and

female part-time workers in columns (4) and (5). However, none of these characteristics is significantly affected

by the reform. We find some (marginally) significant coefficients in Panel A in the specification exploiting short

term variation in outcomes among listed firms. For example, the average share of female employees increases by

1.3 percentage points from a mean of 41%. But these effects are not robust over the longer term horizon in the

comparison with matched control firms. In panel B, all estimated coefficients are very small and statistically

insignificant.

We have seen in Table 3 that companies which have a relatively high share of female board members already in

2012 tend to be more diverse in terms of leadership and also in terms of overall gender workforce composition.

It is thus interesting to see whether firms that are mandated a large adjustment in the number of female board

members, become more similar to companies that started out with higher shares of female directors. In Table 7

we examine heterogeneity in the effects by the incentives for adjustment due to the Golfo-Mosca law. As a

robustness check, Online Appendix Table A5 shows the estimates with a modified comparison group, where we

match a separate comparison sample to each of the three groups of firms. The estimation results do not confirm

our initial hypothesis. If anything, low incentive firms who already have relatively more women on their boards

make the biggest changes in female leadership positions after the reform. Column (3) in Table 7 shows

additional female CEOs are exclusively appointed in the low incentive firms, while there are no adjustments in

medium and high incentive firms. The point estimates imply a 100% increase in female CEOs among low

incentive firms. With respect to the increase of females in the top quartile and deciles of the earnings

distribution, the effects are also most pronounced and robust in the low incentive group, but they are

quantitatively less important. In addition, there is some indication of positive effects on the probability of having

16

a female manager and the share of female managers among medium incentive firms, but estimates are small and

not robust across specifications.

This result is counter intuitive, because it is not clear why low incentive firms would change the gender

composition in top positions in lock step with the timing of board elections. It is more plausible that firms who

started the gender equalization process earlier moved to a differential trend with respect to the gender diversity

measures and thus the common trend assumption is violated in the heterogeneity analysis. Unfortunately, our

sample of listed firms is too small to perform rigorous tests for common trends in the matched subsamples. We

should thus interpret the estimated effects on female CEOs with caution.

6. Conclusion

Over the last decade, many European countries have followed the Norwegian example and introduced laws

mandating gender quota for corporate board membership. In this paper we have evaluated an Italian reform

linking firm-level information on board membership and board renewals with detailed employment and earnings

information from INPS archives. As in other countries, Italian listed companies complied swiftly with the gender

quota once the law was implemented. From 2011 to 2017 the number of board seats taken by women increased

four-fold. We find evidence that companies appreciated the gradual introduction of the quota regulation as some

of them responded in anticipation of the law and started raising their gender quotas at board elections scheduled

in 2012.

Taking advantage of the staggered introduction of the gender quota and variation in board renewals across firms,

we evaluate the effect of the sharp increase of female representation on boards of directors on several measures

of gender gaps in the workforce. Our results provide no evidence that the female board quota mandated by the

reform translated into an increase in female representation at the top executive level or among top earners, at

least not in the short run. There is some indication that listed companies promoted one of their female managers

as a CEO. But these promotions did not lead to a higher representation of women among top earners in the

company. In addition, heterogeneity analysis by the pre-reform share of female board members reveals that new

appointments of female CEOs occurred only in companies that already fulfilled the female quota in 2012.

Furthermore we find no evidence of changes in the overall gender workforce composition due to the reform.

Our findings are in line with evidence for Norway by Bertrand et al. (2019). Even though Italy is a much less

gender egalitarian country than Norway with lower female labor force participation, higher gender wage gaps,

and lower representation of women in the top of the earnings distribution, the strict enforcement of gender ratios

on boards of directors do not have a trickle-down effect that leads to changes in the overall labor market

situation of women.

There are several potential explanation for the absence of effects from the Italian board quota. First, the number

of high profile positions created by the reform is relatively limited. In 2017, when the law was almost fully

17

implemented, 758 director position in Italy were filled by women. Compared to the overall market of highly

qualified women in the country this seems to be a symbolic number which suggests that the coverage of the law

should be extended to a much wider range of companies. Second, if the law affects perceptions and social norms

about women in top positions, economic outcomes might respond with some delay and our analysis of short-

term effects might not capture the full extent of its impact. If norms respond very slowly, it puts into question the

temporary nature of the Golfo-Mosca law which is scheduled to expire after the third board election. Third,

newly appointed female board members might not be in powerful positions that allow them to influence changes

at the firm level. In a recent study Rebérioux and Roudaut (2016) show that newly appointed females on boards

of French companies are less likely to hold key positions than their male counterparts, which weakens any

potential positive effects of gender quota.

In conclusion, while a higher female representation on corporate boards is certainly desirable on the ground of

equity concern, our findings do not support the idea that gender quota alone represents an effective tool to reduce

gender disparities within firms, especially in a country like Italy characterized by a traditional gender culture.

18

References

Adams, R. and Ferreira D. (2009) Women in the boardroom and their impact on governance and performance. Journal of Financial Economics, 94:291–309. Adams, R. and Funk, P. (2012) Beyond the Glass Ceiling: Does Gender Matter? Management Science 58(2): 219-235. Ahern K. R. and Dittmar A. K. (2012) The changing of the boards: the impact on firm valuation of female board representation. The Quarterly Journal of Economics, 127:137– 197. Bagues, M., M. Sylos Labini and N. Zinovyeva. (2017) Does the Gender Composition of Scientific Committees Matter? American Economic Review 107(4): 1207-38. Bertrand, M., and A. Schoar, (2003) “Managing with style: The effect of managers on firm policies”, Quarterly Journal of Economics 118, 1169–1208. Bertrand M., Black S. E., Jensen S. and Lleras-Muney A. (2019) “Breaking the Glass Ceiling? The Effect of Board Quotas on Female Labour Market Outcomes in Norway”, The Review of Economic Studies 86(1), 191-239. Borsa Italiana (2018) https://www.borsaitaliana.it/borsaitaliana/statistiche/mediaitaliano/statistiche/ mercatoprimario/2018/ revoche.en_pdf.htm Cardoso, Ana Rute, and Rudolf Winter-Ebmer (2010) ‘‘Mentoring and Segregation: Female-Led Firms and Gender Wage Policies,’’ Industrial and Labor Relations Review 64, 143–163. Comi, Simona, Mara Grasseni, Federica Origo, and Laura Pagani (2017) “Where Women Make the Difference. The Effects of Corporate Board Gender-Quotas on Firms’ Performance across Europe.” DEMS Working Paper, University of Milan. CONSOB (2017) “Report on corporate governance of Italian listed companies.”, Commissione Nationale per le Società e la Borsa, http://www.consob.it/web/consob-and-its-activities/rcg2017 European Union, 2016. “Gender balance on corporate boards. Europe is cracking the glass ceiling,” Directorate-General for Justice and Consumers, Fact sheet, July 2016. Devicienti Francesco, E.Grinza , A. Manello, D. Vannoni (2019), “What are the benefits of having more female leaders? Evidence from the use of part-time work in Italy”, Industrial and Labor Relation Review, forthcoming. Ferrari, G., V. Ferraro, P. Profeta, and C. Pronzato, (2018). “Gender Quotas: Challenging the Boards, Performance, and the Stock Market,” IZA Discussion Paper No. 10239, September. Ferreira, D. (2015) “Board diversity: Should we trust research to inform policy?” Corporate Governance: An International Review 23: 10 8 –111. Flabbi, L., Macis, M., Moro, A. and Schivardi, F. (2018) Do female executives make a difference? The impact of female leadership on gender gaps and firm performance, Working Paper. Flabbi L. C. Piras and S. Abrahams (2017) “Female Corporate Leadership in Latin America and the Caribbean Region: Representation and Firm-level Outcomes", International Journal of Manpower, 38(6): 790-818

19

Fortin, Nicole M., Brian Bell and Michael Böhm (2017) “Top earnings inequality and the gender pay gap: Canada, Sweden, and the United Kingdom”, Labour Economics, vol. 47, issue C, 107-123. Gagliarducci, S. and Paserman, D. (2014) “The Effect of Female Leadership on Establishment and Employee Outcomes: Evidence from Linked Employer-Employee Data”, IZA DP 8647. Kunze, Astrid, and Amalia R. Miller, (2017) ‘‘Women Helping Women? Evidence from Private Sector Data on Workplace Hierarchies,’’ Review of Economics and Statistics, 99(5): 769–775. Lucifora, Claudio and Daria Vigani (2016) What If Your Boss is a Woman? Work Organization, Work-Life Balance and Gender Discrimination at the Workplace, IZA Discussion Paper 9737. Malmendier, U., Tate and Yan (2011) “Overconfidence and Early-Life Experiences: The Effect of Managerial Traits on Corporate Financial Policies” The Journal of Finance, 76(5), 1687-1733. Matsa, David A., and Amalia R. Miller (2011), Chipping Away at the Glass Ceiling: Gender Spillovers in Corporate Leadership. American Economic Review 101, 635–639. Matsa, David A., and Amalia R. Miller (2013) A Female Style in Corporate Leadership? Evidence from Quotas, American Economic Journal: Applied Economics 5,136–169. Matsa D. A. and Miller A. R. (2014), Workforce Reductions at Women-Owned Businesses in the United States Industrial and Labor Relations Review, 67(2): 422–452. Rebérioux, A. and G. Roudaut (2016) “Gender Quota inside the Boardroom: Female Directors as New Key Players?” CEPREMAP Working Papers, No 1603. Smith, N. (2018) “Gender quotas on boards of directors”. IZA World of Labor 2018: 7 doi: 10.15185/izawol.7.v2. Staines, Graham L., Carol Tavris, and Toby E. Jayarante (1974). “The Queen Bee syndrome”. Psychology Today, 7 (8), 55–60. Tate, Geoffrey, and Liu Yang (2015) ‘‘Female Leadership and Gender Equity: Evidence from Plant Closure,’’ Journal of Financial Economics117, 77–97.

20

Figure 1: Share of Female Board Members based on CONSOB data

Note: Source CONSOB (2016, 2017)

.

5.9

33.6

20Pe

rcen

t

2008 2009 2010 2011 2012 2013 2014 2015 2016 2017

Share of Female Board Members

21

Figure 2: Listed Companies and Matched Comparison Companies over Time

A B

C

D E

22

F G

I

J K

Notes: Figures show annual mean outcome variables for Listed Companies and Matched Limited Comparison Copanies before and after the implementation of the Golfo-Mosca Law in 2012.

23

Figure 3: Workforce characteristics in listed companies relative to reform year 2012

Note: Coefficient estimates of equation (1) relative to mean of dependent variable in the group of listed

companies in 2012

-.05

0.0

5.1

.15

perc

ent d

iffer

ence

2008 2010 2012 2014 2016

Part-time employed Female Part-time employedFemale employed

A. Women in the Workforce0

.51

1.5

perc

ent d

iffer

ence

2008 2010 2012 2014 2016

Female CEO Share Female ManagerShare Female Top Decile Share Female Top Quartile

B. Women in Top Positions

24

Table 1: Female Representation on Corporate Boards Female Directorships Diverse Board Companies

Year Number of female directors Share

Number of companies with at least one female director

Share

2008 170 5,9 126 43,8 2009 173 6,3 129 46,4 2010 182 6,8 133 49,6 2011 193 7,4 135 51,7 2012 288 11,6 169 66,8 2013 421 17,8 202 83,5 2014 521 22,7 217 91,9 2015 622 27,6 230 98,3 2016 701 31,6 226 99,1 2017 758 33,6 227 99,1

Source: CONSOB data on corporate boards of Italian companies with ordinary shares listed in Borsa Italiana SPA–MTA Stock exchange, CONSOB reports 2017, 2018.

25

Table 2: Descriptive Statistics Unmatched Sample Matched Sample Listed Firms Comparison

Group Listed Firms Comparison Group

(1) (2) (3) (4) Employment: Mean 1459 180 631 626 Median 240 664 294 241 Std. Dev. (5652) (623) (1241) (1375) Share of Workers Employed as Apprentice 0,01 0,02 0,01 0,01 (0,05) (0,05) (0,04) (0,02) Blue Collar 0,17 0,39 0,18 0,22 (0,24) (0,31) (0,24) (0,27) White Collar 0,73 0,54 0,74 0,72 (0,25) (0,28) (0,24) (0,26) Manager 0,09 0,06 0,07 0,06 (0,16) (0,13) (0,12) (0,08) Average weekly wage 1076, 474 763,05 1026,20 904,81 (672,99) (500,88) (559,11) (375,26) Wage gap male female 0,68 0,80 0,68 0,72 (0,20) (0,26) (0,18) (0,18) Region: North 73,40 75,02 74,25 79,86 Center 19,68 15,56 17,96 14,09 South 6,92 9,42 7,79 6,05 Industry: Services 54,26 38,10 51,50 51,68 Manufacturing 45,74 61,90 48,50 48,32 Gender diversity Share Female 0,40 0,32 0,41 0,37 (0,18) (0,21) (0,18) (0,19) Share Part Time 0,07 0,07 0,08 0,06 (0,08) (0,10) (0,09) (0,08) Share Female and Part Time 0,06 0,06 0,07 0,06 (0,08) (0,08) (0,08) (0,07) Share Female and Hired 0,04 0,03 0,04 0,03 (0,06) (0,09) (0,06) (0,06) Share Female and Fired 0,03 0,03 0,03 0,03 (0,07) (0,08) (0,06) (0,05) Female ceo 0,04 0,08 0,05 0,05 (0,20) (0,26) (0,21) (0,21) Share Female and Manager 0,02 0,01 0,02 0,01 (0,03) (0,05) (0,03) (0,02) Share Female among Managers 0,18 0,17 0,23 0,16 (0,28) (0,37) (0,13) (0,16) At least one Female Manager 0,71 0,28 0,74 0,66 (0,46) (0,45) (0,44) (0,48) Share Female and Top Quartile Earnings 0,05 0,04 0,05 0,04 (0,04) (0,04) (0,04) (0,04) Share Female and Top Decile Earnings 0,01 0,01 0,01 0,01 (0,01) (0,02) (0,01) (0,01) Number of Companies 188 6338 167 149 Notes: The sample in column (1) includes all companies continuously listed between 2012 and 2016 with matched INPS records in the years 2008 - 2016. Column (2) includes limited companies with at least one manager and continuous INPS records from 2008 - 2016. The matching procedure to select the matched samples in columns (3) and (4) is explained in Section 4. Wage gap is defined as the difference between the average weekly female and male full time equivalent wage.

26

Table 3: Characteristics by Year of Board Renewal

Year of Board renewal 2013 2014 2015

Share Female Directors 2012

8,6 11,2 13,6 (8,5) (8,6) (10,1) Share Female Directors at Renewal 25,4 26,6 28,9 Low Incentive Company 0,17 0,19 0,32 (6,7) (5,9) (7,5) Employment 557 888 507 (953) (1837) (852) Share of Blue Collar 0,136 0,252 0,169 (0,230) (0,256) (0,237) Share of White collar 0,762 0,679 0,755 (0,232) (0,231) (0,240) Share of managers 0,088 0,054 0,067 (0,147) (0,081) (0,115) Share Part Time 0,062 0,080 0,083 (0,056) (0,107) (0,091) Share of female 0,426 0,374 0,414 (0,177) (0,189) (0,180) Share of female part time 0,057 0,073 0,074 (0,055) (0,091) (0,084) Share of female manager 0,020 0,006 0,022 (0,031) (0,008) (0,043) Female CEO 0,052 0,043 0,048

(0,223) (0,204) (0,216) Share Female and Top Decile 0,014 0,013 0,014 (0,015) (0,015) (0,013)

Number of companies 58 47 62 Notes: Low incentive companies are defined as having at least 20% female board members in 2012.

27

Table 4: Descriptives by Type of Reform Incentive Full Sample Low Incentive Medium Incentive High Incentive Listed Comparison Listed Comparison Listed Comparison Listed Comparison (1) (2) (3) (4) (5) (6) (7) (8) Employment: Mean 644 671 490 601 540 561 784 774 Std. Dev. 1291 1510 810 1183 897 1113 1645 1795 Median 291 218 242 268 197 208 291 226 Share of Workers Employed as Apprentice 0,013 0,007 0,008 0,008 0,003 0,002 0,022 0,010 (0,045) (0,025) (0,018) (0,020) (0,010) (0,008) (0,063) (0,033) Blue Collar 0,192 0,209 0,161 0,211 0,182 0,173 0,212 0,227 (0,249) (0,261) (0,254) (0,269) (0,248) (0,244) (0,250) (0,263) White Collar 0,732 0,723 0,770 0,731 0,729 0,753 0,715 0,708 (0,236) (0,244) (0,241) (0,245) (0,235) (0,243) (0,235) (0,241) Manager 0,064 0,061 0,061 0,050 0,086 0,073 0,050 0,055 (0,099) (0,085) (0,085) (0,056) (0,136) (0,111) (0,070) (0,071) Average weekly wage 988,9 927,4 923,5 861,3 1121,8 977,6 931,0 918,9 (517,2) (357,5) (376,9) (242,6) (721,6) (407,7) (385,0) (348,3) Wage gap male female 69,15 71,55 0,68 0,742 0,68 0,71 0,70 0,72 (17,590) (17,600) (0,165) 0.177 (0,182) (0,166) (0,178) (0,179) Region: North 75,5 79,4 66,7 83,3 83,0 75,0 74,7 81,2 Center 14,2 14,2 21,2 6,7 10,6 18,8 16,9 14,5 South 8,6 6,4 12,1 10,0 6,4 6,3 8,5 4,4 Industry: Services 49,7 49,7 46,9 48,4 55,3 54,0 46,5 46,4 Manufacturing 50,3 50,4 53,1 51,6 44,7 45,8 53,5 53,6 Gender diversity Share Female 0,400 0,363 0,454 0,378 0,406 0,376 0,370 0,340 (0,184) (0,183) (0,174) (0,178) (0,179) (0,170) (0,189) (0,192) Share Part Time 0,075 0,060 0,103 0,060 0,071 0,053 0,065 0,063 (0,087) (0,071) (0,112) (0,048) (0,075) (0,050) (0,080) (0,088) Share Female and Part T. 0,068 0,053 0,090 0,053 0,066 0,048 0,058 0,055 (0,078) (0,059) (0,107) (0,046) (0,067) (0,044) (0,068) (0,071) Share Female and Hired 0,037 0,020 0,033 0,025 0,041 0,019 0,036 0,017 (0,065) (0,036) (0,040) (0,031) (0,078) (0,040) (0,066) (0,034) Share Female and Fired 0,030 0,021 0,024 0,017 0,037 0,021 0,028 0,021 (0,057) (0,045) (0,039) (0,023) (0,074) (0,041) (0,051) (0,053) Female CEO 0,053 0,035 0,091 0,067 0,106 0,063 0,000 0,000 (0,225) (0,186) (0,292) (0,254) (0,312) (0,245) (0,000) (0,000) At least one Female Manager 0,748 0,674 0,788 0,700 0,745 0,688 0,732 0,681 (0,435) (0,471) (0,415) (0,466) (0,441) (0,468) (0,446) (0,469) Share Female and Manager 0,017 0,010 0,020 0,011 0,023 0,012 0,011 0,008 (0,034) (0,021) (0,032) (0,017) (0,049) (0,025) (0,018) (0,018) Share Female among Managers 0,263 0,168 0,333 0,216 0,272 0,164 0,213 0,149 (0,182) (0,144) (0,132) (0,166) (0,199) (0,134) (0,112) (0,156) Share Female and Top Quartile 0,050 0,041 0,057 0,049 0,049 0,044 0,047 0,035 (0,042) (0,038) (0,045) (0,037) (0,043) (0,033) (0,040) (0,039) Share Female and Top Decile 0,014 0,012 0,014 0,013 0,015 0,012 0,012 0,010 (0,032) (0,028) (0,036) (0,027) (0,032) (0,022) (0,031) (0,030) Number of Companies 151 141 33 30 47 48 71 69 Notes: Low Incentive means company has at least 20% female directors in 2012; medium Incentive company has between 10 and 20% female directors, high incentive company has less than 10% female directors in 2012. Wage gap is defined as the difference between the average weekly female and male full time equivalent wage. Comparison sample selected by a separate match for each treatment group.

28

Table 5: Estimated Reform Effects on Females in Top Positions

At least one Female Manager

Share Female and Manager CEO is Female

Share Female and Top Quartile

Share Female and Top Decile

(1) (2) (3) (4) (5) A: Listed Firms Reform 0,56 0,04 1,19 0,31 0,06 (2,92) (0,04) (2,81) (0,13) (0,05) Constant 67,00 14,00 -5,01 -0,92 -0,37 (26,40) (2,47) (11,30) (2,86) (0,56) Pre-Reform Mean Dependent Variable 73,70 1,60 4,80 4,90 1,30 R-squared 0.027 0.427 0.008 0.187 0.170 B: Listed Firms and Matched Comparison Firms Reform 3,29 0,00 2,90 0,15 0,06 (2,24) (0,03) (1,23) (0,13) (0,04) Constant 53,70 9,50 -2,55 -0,23 -0,10 (23,00) (3,26) (10,20) (2,49) (0,56) R-squared 0.012 0.277 0.005 0.191 0.136 Notes: Coefficient estimates multiplied by 100; standard errors (clustered by company) in parentheses; regressions include controls for firm size, firm size squared, firm fixed effects and year fixed effects. Numbers of observation: Panel A 1,503 annual observation of 167 listed firms, Panel B: 2844 annual observations of 316 listed and matched comparison firms. Share Female and Manager refers to the company level share of workers who are female and manager. Share Female and Top Quartile (Decile) refers to the share of workers who are female with earnings in the top quartile (decile) of the company specific earnings distribution.

29

Table 6: Estimated Reform Effects on Workforce Characteristics

Share Female Share Female and Hired

Share Female and Fired Share Part Time

Share Female and Part Time

(1) (2) (3) (4) (5) A: Listed Firms Reform 1,30 0,28 0,64 -0,67 -0,58 (0,63) (0,40) (0,38) (0,40) (0,31) Constant 24,80 12,30 4,85 1,50 0,42 (35,60) (6,65) (3,46) (3,94) (3,69) Pre-Reform Mean Dependent Variable 40,70 3,70 3,10 7,50 6,80

R-squared 0.197 0.192 0.071 0.139 0.118 B: Listed Firms and Matched Comparison Firms Reform 0,86 -0,49 -0,11 0,43 -0,02 (0,64) (0,43) (0,17) (0,38) (0,24) Constant 41,60 15,20 5,39 5,96 3,11 (31,00) (6,27) (3,00) (4,46) (3,65)

R-squared 0.154 0.179 0.030 0.099 0.110 Notes: Coefficient estimates multiplied by 100; standard errors (clustered by company) in parentheses; regressions include controls for firm size and firm size squared, firm fixed effects and year fixed effects. Numbers of observation: Panel A 1,503 annual observation of 167 listed firms, Panel B: 2844 annual observations on 316 listed and matched comparison firms.

30

Table 7: Estimated Effects by Reform Incentive

At least one Female

Manager Share Female and Manager

CEO is Female

Share Female and Top Quartile

Share Female and Top Decile

(1) (2) (3) (4) (5) A: Listed Firms Reform Low Incentive -2,29 0,00 7,66 0,59 0,13 (3,37) (0,08) (4,53) (0,19) (0,07) Reform Med. Incentive 8,60 0,09 -1,37 0,03 0,05 (5,60) (0,06) (3,57) (0,18) (0,06) Reform High Incentive -1,55 0,04 -0,34 0,32 0,05 (3,21) (0,05) (3,40) (0,14) (0,06) Constant 67,00 14,00 -8,90 -1,04 -0,41 (25,70) (2,46) (11,00) (2,80) (0,55) R-squared 1,503 0.427 0.021 0.198 0.172 B: Listed Firms and Matched Comparison Firms Reform Low Incentive 0,50 0,05 9,55 0,44 0,13 (2,18) (0,08) (5,05) (0,18) (0,06) Reform Med. Incentive 11,30 0,12 0,49 -0,13 0,05 (5,80) (0,07) (1,84) (0,17) (0,06) Reform High Incentive 1,05 0,07 1,46 0,16 0,04

(2,00) (0,06) (1,11) (0,17) (0,06) Constant 51,20 11,60 -3,57 -1,07 -0,41 (22,40) (2,66) (9,81) (2,47) (0,55) R-squared 0.021 0.307 0.013 0.158 0.136 Notes: Coefficient estimates multiplied by 100; standard errors (clustered by company) in parentheses; regressions include controls for firm size and firm size squared, firm fixed effects and year fixed effects. Numbers of observation: Panel A 1,503 annual observation of 167 listed firms, Panel B: 2844 annual observations on 316 listed and matched comparison firms. Share Female and Manager refers to the company level share of workers who are female and manager. Share Female and Top Quartile (Decile) refers to the share of workers who are female with earnings in the top quantile (decile) of the company specific earnings distribution.

Appendix

Estimation of the propensity score:

We estimate a propensity score for the probability that the firm is subject to the law based on the following

observed workforce characteristics in 2012 and their relative trend over the period 2008-2012: share of female

manager, female CEO probability, probability to observe at least one female among managers, share of part

time, share of female part time, share of females, share of female above the 90th , log of firm size, log of

firmsize squared, share of permanent workers, share of apprentship, white collar, blue collar, and manager,

log of the firm specific total wage cost, log of the firm specific total wage cost squared, share of female above

the 75th and below the 90th percentile of the specific firm wage distribution, share of female above the 90th

of the specific firm wage distribution, female hiring and firing rate, female hiring and firing gross rate,

detailed set of regional (42 provinces available in our sample) and industry (46 industry codes at two digit

available in our sample according to ATECO07) indicators.

We select the matched comparison group of non-listed companies via nearest neighbor matching using a

Mahalanobis matching metric based on the following key variables: share of female managers, female CEO

probability, probability to observe at least one female among managers, share of part time and relative trend,

share of female part time and relative trend, share of females, share of female above the 90th percentile of the

specific firm wage distribution and relative trend, share of female above the 75th and below the 90th percentile

of the specific firm wage distribution and relative trend, log of firm size, log of firmsize squared, share of

permanent workers, share of apprentship, white collar, blue collar, and manager, log of the firm specific total

wage cost, female hiring rate.

Figure A1 shows the distribution of predicted propensity scores among listed companies and control

companies. Table A1 shows the t-tests for the mean difference between listed and matched control group of

all variables and relative trend we use as dependendent variables in our regression model and all variables

included in the matching procedure. The t tests for the mean difference between listed and matched control

group of almost all variables included in the matching procedure are not statistically different from zero.

Figure A1. Distribution of predicted propensity scores

TABLE A1: t test for the mean differences between treated (listed) and control(limited) group Variables Treated Control t p>|T| one female manager 0,7365 0,6577 1,53 0,128 d1 one female manager 0,0060 0,0537 -1,74 0,083 d2 one female manager -0,0120 0,0269 -0,99 0,322 d3 one female manager 0,0000 0,0067 -0,11 0,915 d4 one female manager 0,0539 -0,0067 0,55 0,58 female_ceo 0,0479 0,0470 0,04 0,969 d1 female_ceo -0,0120 0,0134 -1,16 0,249 d2 female_ceo -0,0120 0,0201 -0,97 0,331 d3 female_ceo -0,0240 0,0067 -0,52 0,605 d4 female_ceo -0,0539 -0,0336 -0,18 0,861 share of female manager 0,0165 0,0087 2,49 0,013 d1 share of female manager 0,0001 0,0004 -0,31 0,76 d2 share of female manager -0,0006 -0,0004 -0,08 0,938 d3 share of female manager -0,0011 -0,0004 -0,17 0,866 d4 share of female manager -0,0005 0,0008 -0,22 0,822 75 pct< fem < 90pct 0,0358 0,0315 1,26 0,209 d1 75 pct< fem < 90pct -0,0006 -0,0008 0,16 0,876 d2 75 pct< fem < 90pct -0,0013 -0,0013 0,01 0,994 d3 fem>75 pct< fem < 90pct 0,0006 -0,0006 0,26 0,792 d4 75 pct< fem < 90pct 0,0055 0,0009 0,5 0,615 fem> 75 pct 0,0492 0,0440 1,13 0,261 d1 fem> 75 pct 0,0001 0,0000 0,04 0,968 d2 fem> 75 pct -0,0010 -0,0002 -0,37 0,715 d3 fem> 75 pct 0,0010 0,0013 -0,07 0,947 d4 fem> 75 pct 0,0064 0,0031 0,35 0,725 fem> 90 pct 0,0134 0,0125 0,59 0,558 d1fem> 90 pct 0,0006 0,0008 -0,25 0,801 d2fem> 90 pct 0,0003 0,0011 -0,84 0,401 d3fem> 90 pct 0,0004 0,0019 -0,79 0,431 d4fem> 90 pct 0,0009 0,0022 -0,41 0,679 mean female 0,4067 0,3713 1,7 0,091 d1 mean female 0,0028 0,0036 -0,15 0,88 d2 mean female -0,0051 0,0005 -0,83 0,404 d3 mean female -0,0172 -0,0041 -1,23 0,218 d4 mean female -0,0404 -0,0175 -1,18 0,241 part time 0,0748 0,0644 1,13 0,26 d1 part time 0,0053 0,0037 0,53 0,599 d2 part time 0,0017 0,0023 -0,19 0,852 d3 part time -0,0013 -0,0031 0,25 0,802 d4 part time -0,0113 -0,0223 0,49 0,622 female part time 0,0676 0,0573 1,25 0,212 d1 female part time 0,0053 0,0038 0,56 0,575 d2 female part time 0,0013 0,0028 -0,46 0,644 d3 female part time -0,0024 -0,0004 -0,31 0,755 d4 female part time -0,0136 -0,0134 -0,01 0,993

female hiring rate 0,0372 0,0248 1,76 0,08 d1 female hiring rate -0,0007 -0,0050 0,91 0,363 d2 female hiring rate -0,0010 -0,0095 1,07 0,286 d3 female hiring rate -0,0026 -0,0184 1,08 0,281 d4 female hiring rate 0,0109 -0,0160 0,92 0,36 female firing rate 0,0308 0,0260 0,75 0,454 d1 female firing rate 0,0035 0,0069 -0,63 0,531 d2 female firing rate 0,0082 0,0079 0,04 0,971 d3 female firing rate 0,0102 0,0060 0,25 0,802 d4 female firing rate 0,0089 0,0015 0,25 0,805 ln employment 5,2796 5,3070 -0,15 0,882 ln employment squared 30,7210 30,5900 0,07 0,946 Ln total wage cost 16,09 16,032 0,34 0,734 apprentship 0,0120 0,0065 1,4 0,164 white collar 0,1811 0,2201 -1,35 0,179 blue collar 0,7364 0,7176 0,68 0,497 managers 0,0705 0,0559 1,26 0,208 share of permanent workers 0,9365 0,9476 -1,13 0,258 d1 share of permanent workers -0,0004 0,0034 -0,91 0,364 d2 share of permanent workers -0,0030 0,0056 -1,25 0,212 d3 share of permanent workers -0,0044 0,0091 -1,08 0,282 d4 share of permanent workers -0,0115 0,0077 -0,86 0,39 North 0,7425 0,7987 -1,18 0,239 Center 0,1796 0,1409 0,93 0,352 South 0,0778 0,0604 0,61 0,545 Manufacturing 0,4850 0,4832 0,03 0,974

Table A2: Differences between listed and non-listed firm over time

VARIABLES Part time Female

part time Fem's

wage> 90th

Fem's wage >75th pct <90th

oct

Fem's wage> 75th

Mean Females

Fem hirings

Fem firings

Share of female

Managers

Female CEO

One Female Manager

2009 0.004 0.0025 -7.83e-06 0.000110 0.000103 0.00122 -0.038 0.0018 -0.00133 -0.0087 0.0382

(0.006) (0.004) (0.00067) (0.00148) (0.00181) (0.0069) (0.023) (0.026) (0.00096) (0.024) (0.0586)

2010 0.00149 0.0012 0.000178 -0.00159 -0.00142 -0.0002 0.0159 0.0070 -0.000441 0.0117 0.0762

(0.0041) (0.003) (0.00048) (0.00140) (0.00160) (0.0056) (0.019) (0.026) (0.00077) (0.024) (0.0568)

2011 -0.002 -0.0002 -0.000184 -0.00107 -0.00125 -0.0001 -0.011 0.0072 -0.000470 -0.0109 0.0701

(0.003) (0.002) (0.00052) (0.00116) (0.00135) (0.0049) (0.021) (0.020) (0.00072) (0.023) (0.0540)

2012 0.00196 0.0006 -0.001*** 0.000359 -0.000828 -0.00036 0.0079 0.0059 -0.000737 -0.0218 0.0538

(0.0020) (0.002) (0.00037) (0.00071) (0.00077) (0.0044) (0.02) (0.02) (0.00068) (0.015) (0.0537)

2013 0.00684 0.0028 6.56e-07 -0.000694 -0.000694 0.00360 -0.0259 -0.0140 -0.000250 0.0189 0.0703

(0.0062) (0.004) (0.00084) (0.00184) (0.00234) (0.0103) (0.029) (0.032) (0.00125) (0.023) (0.0620)

2014 0.00847 0.0037 0.000488 -0.000532 -4.40e-05 0.00755 0.0035 0.0002 -0.000173 0.0263 0.0813

(0.0058) (0.004) (0.00081) (0.00209) (0.00261) (0.0098) (0.031) (0.029) (0.00101) (0.0267) (0.0627)

2015 0.00987 0.0028 0.000146 7.64e-05 0.000222 0.00517 -0.0266 -0.0152 0.000145 0.0230 0.0838

(0.0063) (0.004) (0.00092) (0.00203) (0.00271) (0.00983) (0.0304) (0.027) (0.00112) (0.0254) (0.0627)

2016 0.0113* 0.0032 0.000322 -0.000716 -0.000393 0.00357 0.0428 -0.0228 0.000387 0.0374 0.104

(0.00664) (0.004) (0.00074) (0.00219) (0.00262) (0.0108) (0.0331) (0.035) (0.00119) (0.0284) (0.0639)

Const. 0.13*** 0.108*** 0.0178*** 0.0425*** 0.0603*** 0.390*** 0.37*** 0.363*** 0.0066*** 0.0122 0.849***

(0.0039) (0.003) (0.00045) (0.00140) (0.00166) (0.0048) (0.0198) (0.021) (0.0008) (0.0139) (0.0384)

Obs. 2,844 2,844 2,844 2,844 2,844 2,844 2,844 2,844 2,844 2,844 2,844

R-squared 0.099 0.100 0.123 0.092 0.151 0.046 0.020 0.018 0.008 0.006 0.015

Number of group 316 316 316 316 316 316 316 316 316 316 316

Year fixed effect

Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes

Table A3: Placebo test , listed sample

(1) (2) (3) (4) (5) (6)

VARIABLES Part time Part time Mean Female

Female 75=<pct<90

Female pct> =75

Female firing rate

Reform lag1 -0,0004 -0,0012 0,0187*** 0,00316*** 0,00381*** 0,0018

(0,0292) (0,00254) (0,007) (0,000948) (0,00114) (0,0024) Constant 0,0094 -0,0001 0,2470 -0,0055 -0,0090 0,0268***

(0,0391) (0,0371) (0,353) (0,0257) (0,0285) (0,00365)

Observations 1,503 1,503 1,503 1,503 1,503 1,503 R-squared 0.135 0.114 0.202 0.133 0.190 0.062 Number of group 167 167 167 167 167 167

Year fixed effect Yes Yes Yes Yes Yes Yes

Reform lag2 0,01 0,00387 0,00123 -0,00025 -0,00078 -0,00014

(0,00524) (0,00358) (0,00387) (0,00070) (0,00089) (0,00243) Constant 0,117*** 0,0950*** 0,385*** 0,0421*** 0,0603*** 0,0288***

(0,00556) (0,00357) (0,00592) (0,000899) (0,0011) (0,00224)

Observations 1,503 1,503 1,503 1,503 1,503 1,503 R-squared 0.119 0.094 0.040 0.073 0.133 0.061 Number of group 167 167 167 167 167 167

Year fixed effect Yes Yes Yes Yes Yes Yes

Note : Placebo regression for listed firm sample

Table A3_bis : Placebo test , matched sample

CEO is Female CEO is Female

Reform lag1 0,0305** Reform lag2 0,0284

(0,015) (0,0191) Constant 0,000646 Constant -0,00534

(0,0986) (0,105)

Observations 2,844 Observations 2,844

R-squared 0,005 R-squared 0,005

Number of group 316 Number of group 316

Year fixed effect Year fixed effect Yes

Table A4: Estimates Reform Effects on Female Top Positions

At least one Female Manager

Share Female and Manager CEO is Female

Share Female and Top Quartile

Share Female and Top Decile

(1) (2) (3) (4) (5) A: Listed Firms Reform 0,00576 0,000489 0,0114 0,00336*** 0,000637

(0,0292) (0,0007) (0.0285) (0,00116) (0,0005) Constant 0,931*** 0,00586*** 0,0404 0,0566*** 0,0173***

(0,0275) (0,0009) (0,0353) (0,0014) (0,0006)

R-squared 0,025 0,008 0,006 0,14 0,14

B: Listed Firms and Matched Comparison Firms

Reform 0,033 0,000709 0,0290** 0,00154 0,000589

(0,0223) (0,0006) (0,0123) (0,0014) (0,0005) Constant 0,904*** 0,00624*** 0,0265*** 0,0556*** 0,0165***

(0,0082) (0,0002) (0,0071) (0,0004) (0,0002)

R-squared 0,01 0,007 0,004 0,152 0,121

Notes: Coefficient estimates multiplied by 100; standard errors (clustered by company) in parentheses; regressions include firm fixed effects and year fixed effects. Numbers of observation: Panel A 1,503 annual observation of 167 listed firms, Panel B: 2844 annual observations of 316 listed and matched comparison firms. Share Female and Manager refers to the company level share of workers who are female and manager. Share Female and Top Quartile (Decile) refers to the share of workers who are female with earnings in the top quartile (decile) of the company specific earnings distribution.

Table A5: Estimates by the Reform Incentives –Robustness check.

(1) (2) (3) (4) (5)

VARIABLES Share Female and Manager CEO is Female

At least one Female Manager

Share Female and Top Quartile

Share Female and Top Decile

A: Listed Firms Reform Low Incentive -0,001 0,083 -0,022 0,006*** 0,001

(0,0007) (0,052) (0,0385) (0,002) (0,0008) Reform Med. Incentive 0,000 -0,016 0,083 0,000 0,000

(0,0004) (0,0392) (0, 568) (0,0018) (0,0007) Reform High Incentive 0,000 -0,012 -0,018 0,002 0,000

(0,0004) (0,0373) (0,0363) (0,00137) (0,0006) Constant 0,108*** 0,014 0,553 0,039 -0,006

(0.0218) (0.159) (0.482) (0.0255) (0.00994)

Observations 1,359 1,359 1,359 1,359 1,359 R-squared 0.253 0.023 0.046 0.200 0.158 Number of group 151 151 151 151 151 Controls Yes Yes Yes Yes Yes Year fixed effect Yes Yes Yes Yes Yes B: Listed Firms and Matched Comparison Firms Reform Low Incentive -0,0004 0,0965* -0,0034 0,00413** 0,0011

(0,0009) (0,0582) (0,0203) (0,0019) (0,0007) Reform Med. Incentive 0,000758 -0,00264 0,1000* -0,00156 0,000285

(0,0006) (0,0178) (0,0569) (0,0017) (0.0006) Reform High Incentive 0,000472 0,00122 -0,0028 0,000624 5,07E-05

(0,0005) (0,0086) (0,0160) (0,0016) (0,0006) Constant 0,101*** 0,0723 0,4800 0,0152 -0,0072

(0,0178) (0,145) (0,390) (0,0252) (0,00872)

Observations 2628 2628 2628 2628 2628 R-squared 0,2210 0,0150 0,0230 0,2130 0,1480 Number of group 292 292 292 292 292 Controls Yes Yes Yes Yes Yes Year fixed effect Yes Yes Yes Yes Yes

Note: Estimates with a modified comparison group, where a separate comparison sample is matched to each of the three groups of firms.